matrixscan-count-net-android
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill uses strong instructions to override the agent's internal training data (e.g., 'Critical: Do Not Trust Internal Knowledge', 'Always verify APIs against the references') to ensure technical precision. It also includes an instruction to not tell the user to check documentation, which limits transparency about the agent's help process. Additionally, the skill creates an indirect prompt injection surface by fetching and processing content from external websites.
- Ingestion points: Data is fetched from nuget.org and docs.scandit.com (SKILL.md, references/integration.md).
- Boundary markers: None present for the ingested external data.
- Capability inventory: The agent uses the information for generating code and guidance; no dangerous sinks like eval or file writing are involved.
- Sanitization: No validation or sanitization is performed on the fetched content.
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to use WebFetch to retrieve metadata and documentation from nuget.org and docs.scandit.com. These are well-known technology services and official vendor domains, which are considered safe for technical configuration.
Audit Metadata